Add Preprocessing Code

src_preprocessing/run_abn_preprocess.py

@@ -0,0 +1,59 @@
+import json
+import os
+import argparse
+
+# Run: python src_preprocessing/run_abn_preprocess.py \
+    # --mimic_cxr_annotation data/mimic_cxr_annotation.json \
+        # --mimic_abn_annotation data/mimic_abn_annotation.json \
+            # --image_path data/mimic_cxr_jpg \
+                # --output_path data/mimic_abn_annotation_processed.json
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--mimic_cxr_annotation", type=str)
+parser.add_argument("--mimic_abn_annotation", type=str)
+parser.add_argument("--image_path", type=str)
+parser.add_argument("--output_path", type=str)
+args = parser.parse_args()
+
+
+ref_anno = json.load(open(args.mimic_cxr_annotation, "r", encoding="utf-8"))
+anno = json.load(open(args.mimic_abn_annotation, "r", encoding="utf-8"))
+
+img2report = {}
+miss_count = 0
+for key in anno:
+    for sample in anno[key]:
+        for img_path in sample["image_path"]:
+            # should be changed according:
+            # Here is an example for the Chen et al. (2020) annotation:
+            # Chen et al. (2020) Generating Radiology Reports via Memory-driven Transformer. In Proceedings of EMNLP 2020.
+            img_path = os.path.join(args.image_path, img_path.replace("files/", ""))
+            img_path = img_path.replace(".dcm", ".jpg")
+            if not os.path.exists(img_path):
+                print("not exist", img_path)
+                miss_count += 1
+                continue
+            image_id = img_path.split("/")[-1].split(".")[0]
+            img2report[image_id] = sample["report"]
+print(len(img2report), miss_count)
+
+ref_ids = {}
+new_anno = {}
+for key in ref_anno:
+    for sample in ref_anno[key]:
+        ref_ids[sample["id"]] = (key, sample)
+print(len(ref_ids))
+
+updated_anno = {}
+miss_count = 0
+for img_id in img2report:
+    if img_id not in ref_ids:
+        continue
+    if ref_ids[img_id][0] not in updated_anno:
+        updated_anno[ref_ids[img_id][0]] = []
+    new_sample = ref_ids[img_id][1]
+    new_sample["report"] = img2report[img_id]
+    updated_anno[ref_ids[img_id][0]].append(new_sample)
+print([len(val) for val in updated_anno.values()])
+with open(args.output_path, "w", encoding="utf-8") as f:
+    json.dump(updated_anno, f, indent=4, ensure_ascii=False)